case study
A New Jersey home-care agency. Getting 45 caregivers’ weekly timesheets and care notes reviewed, official, and filed took the office most of a week. We audited the cycle; Slate — software the agency owns — closes it in two business days.
measured from the systems’ own logs, at the agency’s own rates · the full math lives in the operating map →
the client
the problem
Traced from the systems’ own logs across four weekly cycles. Open any step.
DSPs care for participants at home. Most keep paper notes during visits — time, activities, mileage — to be turned into forms later.
saves ~20 m per worker — recorded during the visit, not reconstructed from memory
Forms filled from the week’s notes — or from memory where notes are thin — and submitted by the Sunday 8 pm deadline. 26% miss it.
the packet checks itself before it submits — nothing incomplete gets through to Monday
The office coordinator collects what arrived — platform submissions, emailed PDFs, paper drops — logs each packet in the tracker, and rebuilds the who’s-missing list from memory and inbox scrolling.
saves 1 h 15 m — all of it: the queue builds itself
Calls and texts to the 18 workers whose packets are missing or incomplete — Monday through Wednesday, between everything else.
saves ~5 h — reminders go out before the deadline; the coordinator calls only the silent few
Hours and mileage typed again into the pay sheet and the billing sheet, then checked against each participant’s authorized weekly hours — kept in a third spreadsheet.
re-keying goes to zero — approved packets export straight into pay & billing
Each packet read against the participant’s service plan and authorized hours. Errors found here are already a week old.
grows 20 m — on purpose: review is the human judgment, and it gets deeper, not thinner
Fifteen packets a week go back to their workers — a missing signature, a thin outcome note, overlapping times — by phone and text. Corrected versions re-enter at step #03.
saves 3 h 45 m — returned in-app with a named reason, not by phone tag
The state-format weekly form rebuilt by hand from the pieces — data pasted in, signature pages attached, a multi-page PDF made — then saved into SharePoint folders one by one and the tracker updated.
saves 3 h 25 m — the PDF generates and files itself on approval; the office spot-checks
across four measured weeks: 24 h 15 m of office hands-on work per cycle · filed 8 days after the week ends (median) · worst on record: 16 days — after the build: ~5 h per cycle, reviewed and filed in 2 business days
One hour with the owner, on a video call — human-led. AI assists; it does not drive the conclusions.
A short summary back in the owner’s language, three days later. The client approves what moves into the audit.
How the paperwork really moves — learned from the people who run it, not from memory.
A human-reviewed report — every finding survived business and technical review before the agency saw it.
the deliverables — as the agency receives them
deliverable 1 · scope confirmation The findings, for approval What we heard, the three processes with what each costs, the recommendation, and the scope choice — two pages. read the 2-page findings → deliverable 2 · audit delivery The master operating map The audit workspace: the findings at a glance, then nine documents in reading order — the evidence, the numbers, the rebuild, the build, the run. open the operating map →the record